feat(abi): repair from the recorded verdicts rather than an ad-hoc scan #206
2026-09-07
feat(abi): repair from the recorded verdicts rather than an ad-hoc scan
The earlier repair worked from package lists produced by a one-off local symbol scan. Those lists were derived from a candidate set seeded on one slot, so they were a lower bound by construction, and they do not survive the machine that made them. abi_flat_safety now holds the authoritative answer for every slot: the flat-safety backfill inspected all 95418 objects and recorded which cannot load under which minor. 4077 are unsafe, against the 93% of compiled packages the source classifier suspects. Each object is rebuilt under every installed R minor with no --sensitive-only gate. Since bincraft v5.3.0 the destination is decided by the built binary's symbols, so a package needing a per-minor slot lands there and a portable one lands in the flat slot. Gating on the source classifier is what let Rcpp slip through and break every R 4.4 client on the slot. This change will: - add local/list-unsafe-flat-objects.R, printing the recorded unsafe objects - fail loudly when a slot has no verdicts at all, since an empty repair list would otherwise read as "nothing to do" - add a manual workflow rebuilding them under every installed minor